Sustained Low Incidence of Severe and Fatal COVID-19 Following Widespread Infection Induced Immunity after the Omicron (BA.1) Dominant in Gauteng, South Africa: An Observational Study

We conducted an epidemiologic survey to determine the seroprevalence of SARS-CoV-2 anti-nucleocapsid (anti-N) and anti-spike (anti-S) protein IgG from 1 March to 11 April 2022 after the BA.1-dominant wave had subsided in South Africa and prior to another wave dominated by the BA.4 and BA.5 (BA.4/BA.5) sub-lineages. We also analysed epidemiologic trends in Gauteng Province for cases, hospitalizations, recorded deaths, and excess deaths were evaluated from the inception of the pandemic through 17 November 2022. Despite only 26.7% (1995/7470) of individuals having received a COVID-19 vaccine, the overall seropositivity for SARS-CoV-2 was 90.9% (95% confidence interval (CI), 90.2 to 91.5) at the end of the BA.1 wave, and 64% (95% CI, 61.8 to 65.9) of individuals were infected during the BA.1-dominant wave. The SARS-CoV-2 infection fatality risk was 16.5–22.3 times lower in the BA.1-dominant wave compared with the pre-BA.1 waves for recorded deaths (0.02% vs. 0.33%) and estimated excess mortality (0.03% vs. 0.67%). Although there are ongoing cases of COVID-19 infections, hospitalization and death, there has not been any meaningful resurgence of COVID-19 since the BA.1-dominant wave despite only 37.8% coverage by at least a single dose of COVID-19 vaccine in Gauteng, South Africa.


Background
By 14 November 2021, over 40% of the global population had been infected at least once by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) [1]. In mid-November 2021, the highly transmissible severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) Omicron BA.1 (B.1.1.529; henceforth referred to as BA.1) variant of concern (VOC) was identified in Southern Africa, heralding a resurgence of COVID-19 globally [2]. The intrinsic transmissibility of BA.1 is estimated to be twice as high as that of the Delta VOC, which has a basic reproduction rate (Ro) of 5-6 [3]. The BA.1 VOC is resistant to the neutralizing activity of antibodies induced by first-generation COVID-19 vaccines and infection by the wild-type virus or earlier VOCs [4]. Nevertheless, CD4 + and CD8 + T-cell immunity against BA.1 induced by prior infection with SARS-CoV-2 ancestry virus and

Statistical Analysis
The anti-N IgG sensitivity for detecting past infection was previously reported as 58.0%; hence, we used anti-N or anti-S IgG positivity to characterise overall seroprevalence [10]. Anti-S IgG positivity in individuals who received a COVID-19 vaccine (either A26.CoV2.S or BNT162.b2 in South Africa) could be due to vaccination or past infection.
The criteria used to determine serological evidence of SARS-CoV-2 infection during the interval between the pre-BA.1 and post-BA.1-dominant wave serosurveys among those individuals with paired samples from each survey are detailed in Supplementary Table  S2a,b. In individuals who had not received any COVID-19 vaccine in the interval between the two serosurveys and who were seronegative for anti-N or anti-S IgG in the earlier study, seropositivity for either anti-S or anti-N IgG was defined as seroconversion, respectively. For individuals who were anti-N or anti-S IgG positive in the previous survey, seroresponse was defined by a two-fold or greater increase in anti-N IgG or anti-S IgG between the two time points, respectively. The calculation of serological evidence of presumed BA.1 infections was based on either seroconversion or seroresponse between the survey time points. In individuals with paired samples who received a COVID-19 vaccine in the interval between the two serosurveys, a two-fold increase in anti-N IgG or seroconversion for anti-N IgG (i.e., negative to positive) was used as serological evidence of SARS-CoV-2 infection.
The percentage of seropositivity for either anti-N or anti-S in the COVID-19-unvaccinated individuals in the survey prior to the BA.1-dominant wave, multiplied by the STATS-SA population, yielded the inferred number of infections over the course of the first three COVID-19 waves prior to BA.1. The percentage with serological evidence of infection (composite of seroresponse and seroconversion) in the COVID-19 unvaccinated individuals with paired samples multiplied by the STATS-SA population [14] yielded the inferred number of infections during the BA.1 dominant wave. Inferred numbers of SARS-CoV-2 infections were used to calculate ratios of inferred numbers of infections to recorded COVID-19 cases, hospitalizations, and deaths. Inferred numbers of infections at the population level allowed the direct calculation of IFRs, which are the inverse of the inferred infections: recorded (or excess mortality attributable) COVID-19 deaths ratio. Data were analysed using R v4.1.1 (Vienna, Austria) and STATA v16.1 (College Station, TX, USA).

Survey Ethics
The Human Research Ethics Committee at the University of the Witwatersrand granted a waiver for ethics approval of the survey, which was conducted as part of public health surveillance by the Gauteng Department of Health. All participants were, however, required to provide written informed consent, and individuals within a household were free to decline participation.

Participants
We surveyed 3345 households, including 1052 (31.4%) enrolled in the previous survey. Dried blood spots were obtained from 7510 individuals, including 2420 (32.2%) with paired samples, as shown in Figure 1 and Supplementary    We show the flow of participants included in survey analyses from identifying the households and approaching individuals for informed consent through to specimen collection, processing, and data analyses. The final analysis included 7510 individuals in 26 sub-districts. Of note, 4590 individuals from the pre-omicron serosurvey could not be sampled during the post-BA.1-dominant wave because 79 individuals (1.1%) out-migrated, and 37 individuals (0.5%) died between the pre-and post-Omicron BA.1-dominant waves. Furthermore, 726 individuals (10.4%) could not be reached because their households were in estates where access was denied, and 185 refused to participate. Finally, 3563 were unavailable for sampling in the current serosurvey.
Only 29.0% (n = 1995/6886) of individuals older than 12 years of age who were eligible to receive a COVID-19 vaccine had received at least a single dose. Seropositivity was slightly higher in individuals older than 12 years who had received at least a single dose of the COVID-19 vaccine (96.1%; 95% CI: 95.2-96.9) compared with unvaccinated individuals (89.5%; 95% CI: 88.6-90.3). Higher seropositivity was evident for the vaccinated compared with the unvaccinated across all age groups > 12 years (see Table 2) and across districts and sub-districts (see Supplementary Tables S5 and S6). n, number with outcome. N, denominator of population sampled. 1 Of the 2420 paired samples available, 2058 did not receive a COVID-19 vaccination between the pre-and post-Omicron BA.1 serosurveys. 2 Seroprevalence was defined as seropositive for anti-S or anti-N IgG irrespective of vaccination status. 3 Seroconversion is defined for individuals who were seronegative to both S and N on the pre-Omicron serosurvey and seroconverted to either S or N on the post-Omicron BA.1 serosurvey. 4 Seroresponse for anti-N and/or anti-S IgG is defined for individuals who were seropositive to either S or N on the pre-Omicron serosurvey and either seroconverted to S, seroconverted to N, or were seropositive to N on the pre-Omicron BA.1 serosurvey and had a ≥2-fold increase in anti-N titers on the post-BA.1-dominant wave serosurvey or were seropositive to S on the pre-BA.1 serosurvey and had a ≥2-fold increase in anti-S titers on the post-BA.1 serosurvey. 5 Overall serological evidence of SARS-CoV-2 infection in the period between the two surveys when the BA.1-dominant wave occurred was defined as either seroconversion or seroresponse for anti-N and/or anti-S IgG. 6 CI, confidence interval; confidence intervals have not been adjusted for multiplicity and should not be used for inference. The criteria used for determining seroconversion and seroresponse are outlined in Supplementary Table S2b.
In individuals with paired samples who were seropositive during the pre-BA.1 serosurvey and did not receive any COVID-19 vaccine between the two sampling points, 20% (179/896) and 4.2% (65/153%) had seroreversion for anti-N and anti-S IgG seropositivity, respectively. Furthermore, 3.1% (49/1591) of these individuals who were seropositive for either anti-N or anti-S IgG at the pre-BA.1 serosurvey tested seronegative for both epitopes in the latest survey. Whereas the BA.1-dominant wave contributed to 13.8% and 5.8% of the cumulative number of recorded COVID-19 hospitalizations and deaths, respectively, the corresponding percentages for the eight-month period in the Omicron sub-lineages era were 7.2% and 3.0%; Table 3 and Figure 2. The cumulative excess mortality estimates in the post-BA.1 period (n = 6753) was, however, higher than during the BA.1-dominant wave (n = 2974) and contributed to 10.2% and 4.5% of excess deaths due to natural causes since the start of the pandemic, respectively.

COVID-19 Rates, Hospitalizations and Deaths
The cumulative incidence rate (per 100,000) of recorded COVID-19 cases declined from 5957 over the first three COVID-19 waves (pre-BA.1) to 1805 in the BA.1-dominant wave and 523 in the Omicron sub-lineages era. The inferred infections to recorded COVID-19 cases ratio increased from 9.1 in the pre-BA.1 period to 36.3 in the BA.1 dominant wave, indicating greater under-ascertainment of infections in the latter wave; see Table 3 and Figure 2.
The cumulative incidence rate (per 100,000) of COVID-19 hospitalizations declined from 822 in the pre-BA.1 period to 143 and 75 in the BA.1-dominant wave and the Omicron sub-lineages era, respectively. Seven percent of all COVID-19 hospitalizations since the start of the pandemic occurred in the eight-month Omicron sub-lineages era, compared with 13.8% during the BA.1-dominant wave over a five-month period. Whereas an estimated 66 infections resulted in one COVID-19 hospitalization during the pre-BA.1 period, there was one COVID-19 hospitalization for every 458 inferred infections during the BA.1 dominant wave; see Table 3.
cron sub-lineages era, respectively. Seven percent of all COVID-19 hospitalizations since the start of the pandemic occurred in the eight-month Omicron sub-lineages era, compared with 13.8% during the BA.1-dominant wave over a five-month period. Whereas an estimated 66 infections resulted in one COVID-19 hospitalization during the pre-BA.1 period, there was one COVID-19 hospitalization for every 458 inferred infections during the BA.1 dominant wave; see Table 3.    The inferred number of infections in the population pre-Omicron BA.1 dominant wave was derived by multiplying the seroprevalence in unvaccinated individuals at the time of the pre-BA.1 serosurveys by the STATS-SA population [14]. The post-BA.1 inferred number of infections was obtained by multiplying the proportion of unvaccinated individuals showing overall serological evidence of SARS-CoV-2 infection (Table S8) between the pre-BA.1 and post-BA.1-dominant wave serosurveys, by the STATS-SA population. 2 The infection fatality ratio was calculated as the inverse of the inferred infection to recorded deaths or excess ratios. All data are from the National Institute for Communicable Diseases daily databases [12] except for weekly excess deaths. Excess mortality from natural causes was defined per and sourced from the South African Medical Research Council [13]; the excess mortality data are reported through 4 June 2022. Other waves are lagged with respect to cases. Consequently, each of the hospitalization, recorded death, and excess death waves has its own cut-off points determining the start and end of the four epidemic waves. † Changes in testing rates, particularly the lower rates during Wave 1 due to constraints in laboratory capacity and prioritization of testing for hospitalized individuals, prevent direct comparisons, especially in terms of case numbers during the first wave in relation to the subsequent waves. Cases include asymptomatic and symptomatic individuals. Cumulative reported cases were sourced from the National Department of Health. ‡ Hospitalization data are from DATCOV, hosted by the National Institute for Communicable Disease, [12] as described previously [10,11]. The system was developed during the course of the first wave, with gradual onboarding of facilities; hence, these data could underestimate hospitalized cases in the first wave relative to subsequent waves. The hospitalized cases include individuals with COVID-19, as well as coincidental infections identified as part of routine testing for SARS-CoV-2 of individuals admitted to the facilities to assist in triaging of patients in the hospital. § Cumulative reported deaths were sourced from the National Department of Health.
For COVID-19 cases, the pre-BA. All rates were smoothed using a 7-day moving average except for excess mortality. The shaded area in grey indicates when the serosurvey sampling was undertaken.
The cumulative incidence rate (per 100,000) of recorded COVID-19 deaths declined from 180.6 to 11.6 and 5.90 during the pre-BA.1, BA.1-dominant wave and the Omicron sub-lineage era, respectively. Ninety-one percent of all recorded deaths since the start of the pandemic preceded the onset of the BA.1 dominant wave, 5.8% occurred during the BA.1-dominant wave, and only 3.0% occurred in the Omicron sub-lineages era. Overall, there was one recorded COVID-19 death for every 300 and 5643 inferred infections, with IFR of 0.33% and 0.02% in the pre-BA.1 period and BA.1-dominant wave, respectively; see Table 3.
The cumulative incidence rate (per 100,000) of COVID-19-attributable deaths using the excess mortality estimates was approximately 2.2 times higher compared with the recorded deaths (425 vs. 191.7, respectively). The cumulative incidence rate of excess deaths declined from 362.6 over the pre-BA.1 period to 19.2 during the BA.1-dominant wave and was 43.6 in the Omicron sub-lineage era. Overall, there was one death for every 149 and 3719 inferred infections, corresponding to IFRs of 0.67% and 0.03% in the pre-BA.1 period and BA.1-dominant wave, respectively; see Table 3.
An age-group-stratified analysis of the cumulative incidence rates showed the same downward trend of recorded COVID-19 cases, hospitalizations and deaths from the pre-BA.1 period through the BA.1-dominant wave to the Omicron sub-lineage era, as shown in Figure 3A-D and Supplementary Table S10. In adults older than 50 years of age, the BA.1dominant wave and Omicron sub-lineage era, respectively, contributed to 9.1% and 6.0% of COVID-19 hospitalizations and 5.0% and 2.8% of recorded deaths that occurred since the start of the pandemic. Additionally, comparing the pre-BA.1 period with the BA.1-dominant wave, the ratio of inferred infections for every COVID-19 hospitalisation increased from 11.8 to 213 and from 38 to 1352 for recorded COVID-19 deaths, corresponding to IFR values of 2.63% and 0.07% (Supplementary Table S10).

Discussion
Despite only 26.7% (1995/7470) of individuals in the survey having received at least a single dose of COVID-19 vaccine at the time of the serosurvey after the BA.1-dominant wave, the overall seropositivity for SARS-CoV-2 was 90.9% after the BA.1-dominant wave had subsided in Gauteng; including 89.5% in COVID-19 unvaccinated individuals older than 12 years of age. Using paired serology data, 63.9% of the population was infected with the BA.1 variant during the fourth COVID-19 wave in Gauteng. Serological evidence of infection during the BA.1-dominant wave was higher (67.0%) in individuals who were unvaccinated against COVID-19 compared with those who had only been vaccinated before the onset of the BA.1-dominant wave (54.8%). Against this background of high seropositivity and high rates of BA.1 infections, we observed even further decoupling of SARS-CoV-2 infection and recorded COVID-19 hospitalizations and deaths during the subsequent emergence of Omicron sub-lineages. Although these resulted in a modest resurgence of COVID-19 infections in Gauteng and in South Africa, the burden of the disease was even further attenuated compared with the BA.1-dominant wave. Although there has been circulation of newer sub-lineages of Omicron since June 2002 in South Africa, including 22E/BQ.1.*, XAY and XBB, BA.5 has remained the dominant circulating sub-lineages of Omicron [15]. The Omicron sub-lineage era contributed to 7.2% and 3.0% of all recorded COVID-19 hospitalizations and deaths (over an eight-month period) since the start of the pandemic, compared with 13.8% and 5.8% of these events during the BA.1-dominant wave (over a five-month period).
We were also able to provide an update on the burden of COVID-19 during the BA.1-dominant wave compared with the earlier three COVID-19 waves. The SARS-CoV-2 cumulative IFR declined by a factor of 16.5, one order of magnitude, from 0.33% to 0.02% for recorded deaths and by a factor of 22.3 from 0.67% to 0.03% for excess deaths in the pre-BA.1 waves compared with the BA.1-dominant wave. In comparison, the IFR for seasonal influenza virus pre-COVID-19 pandemic in South Africa is estimated to be 0.05% based conservatively on approximately 35% (n = 20.9 million) of the population with serological evidence of infection and excess deaths attributed to influenza of 11,000 per annum [16][17][18]. We may, however, have overestimated the IFR in the pre-BA.1 period, as the cross-sectional seroprevalence used to infer infections over the course of the first three COVID-19 waves would have missed re-infections.
Notably, there has been a 1.93-fold decrease in recorded COVID-19 deaths over the five-month period that constituted the BA.1-dominant wave (n = 1802) compared with the subsequent eight-month period when Omicron sub-lineages have dominated (n = 913). The excess deaths, however, were higher during the latter period (n = 6753) compared with the BA.1-dominant wave (n = 2974). The ratio of excess deaths to recorded deaths increased from 1.65-fold (2974/1802) in the BA.1-dominant wave to 7.40 (6753/913) in the Omicron sub-lineage era. It is unlikely there has been differential under-reporting of COVID-19 deaths between these two periods to explain this discrepancy, as there were no changes in access to healthcare, and DATCOV surveillance has been stable during this time. Rather, these data suggest that excess mortality calculations are becoming less reliable in reflecting the true extent of COVID-19 deaths. This could be due to a number of factors, including the re-emergence of other infectious diseases. The Omicron sub-lineage era in Gauteng coincided with an epidemic of respiratory syncytial virus, as well as an influenza virus epidemic, which occurred earlier than anticipated based on pre-COVID-19 seasonal epidemiology [19]. By contrast, seasonal influenza was largely absent during 2020, and there was only a mild influenza season in 2021 in South Africa. Other chronic causes of death aggravated by the socio-economic fallout of the pandemic and related control measures and reduced effectiveness of the healthcare system during the course of the COVID-19 pandemic could also have contributed to variation over time in the consistency of excess mortality estimates as a proxy for COVID-19 attributable deaths. Other limitations of our study are elaborated upon in the Supplementary Material.
There is uncertainty as to whether the BA.1 variant is intrinsically less virulent than earlier variants [20,21]. The propensity of BA.1 to infect the upper rather than the lower airways could have contributed to the decoupling of infection and COVID-19 [22][23][24]. The BA.2 sub-lineage was, however, associated with high mortality rates in Hong Kong in populations with low levels of immunity. Nevertheless, our study indicates that unless future variants harbour mutations that evade poly-epitopic CD4 + and CD8 + immunity induced by current vaccines and past infection, or the virus becomes intrinsically more virulent, COVID-19 no longer poses a major threat of a large burden of severe disease and death compared with the period prior to late 2021 before the evolution of extensive population immunity. Notably, the breadth of T-cell immunity against SARS-CoV-2 is expected to be more diverse in settings such as South Africa, where there has been a high rate of SARS-CoV-2 infection [25]. Poly-epitopic T-cell responses following SARS-CoV-2 infections by all variants to date are directed against the spike, nucleocapsid and membrane protein epitopes. These responses contribute to attenuating the progression of infection to severe disease and may also reduce transmission of the virus [7]. Although hybrid immunity with three doses of COVID-19 vaccines protects better against BA.1-symptomatic COVID-19 than infection-only induced immunity, both are similarly effective against severe COVID-19 [19].
Following the waning of the BA.4/BA.5-dominant wave, South Africa ended all restrictions aimed at limiting the transmission of SARS-CoV-2 on 22 June 2022 [26]. Nevertheless, recorded COVID-19 cases, hospitalizations and death have not shown significant increases since then despite the emergence of further sub-lineages of Omicron, which are even more antibody-evasive and have a better propensity than BA.1 to infect the lower airways [27][28][29]. Additionally, the absence of any further resurgence of COVID-19 since the end of the BA.4/BA.5-dominant wave has persisted despite the vaccine coverage in Gauteng for at least a single dose of COVID-19 vaccine by 14 November 2022 only being 37.8% (n = 5,905,254) [30].
The experience in South Africa indicates very high rates of infection and re-infection amongst both the unvaccinated and vaccinated decoupled SARS-CoV-2 infections from severe disease in a population with very high seropositivity prevalence, heralding an endemic phase to the pandemic akin to other endemic respiratory viruses that may cause epidemics, such as seasonal influenza. The deployment of COVID-19 vaccines in Africa has lagged behind other regions, with only 32% of the population having received at least a single dose of COVID-19 vaccine as of December 12, 2022 compared with 69% of the overall global population [31]. Nevertheless, our findings indicate that the strategy and target of the current generation of COVID-19 vaccine rollout need to be reconsidered in Africa, where the seropositivity for SARS-CoV-2 was already 86.7% (84.6-88.5%) in December 2021 [32]. Considering that even up to four doses are sub-optimal in sustaining protection against SARS-CoV-2 infection and non-severe COVID-19 [33], the focus in Africa should be targeting high-risk groups who may remain susceptible to developing severe COVID-19 even after past SARS-CoV-2 infection with or without past vaccination.  Table S1: Time period over which serosurvey was undertaken in different sub-districts in Gauteng, South Africa; Table S2a: Vaccination classification of individuals with paired samples for immunoglobulin G analysis at the pre-BA.1 and post-BA.1 sero-surveys; Table S2b: Criteria used to determine seroconversion or seroresponse in the pre-Omicron and post Omicron BA.1 sero-survey interval period; Table S3: Demographic characteristics and anti-nucleocapsid or anti-spike protein IgG seroprevalence in individuals sampled in only post-BA.1 sero-survey and those samples in both pre-BA.1 and post-BA.1 sero-surveys; Table S4: Seroprevalence of SARS-CoV-2 anti-spike (anti-S) or anti-nucleocapsid (anti-N) immunoglobulin G (IgG) Gauteng Province across the Districts and sub-Districts irrespective of COVID-19 vaccination status; Table S5: Seroprevalence of SARS-CoV-2 anti-spike (anti-S) or anti-nucleocapsid (anti-N) immunoglobulin G (IgG) Gauteng Province across the Districts and sub-Districts vaccinated individuals older than 12 years only; Table S6: Seroprevalence of SARS-CoV-2 anti-spike (anti-S) or anti-nucleocapsid (anti-N) immunoglobulin G (IgG) Gauteng Province across the Districts and sub-Districts unvaccinated individuals older than 12 years only; Table S7: Sero-response to SARS-CoV-2 anti-spike (anti-S) and anti-nucleocapsid (anti-N) immunoglobulin G (IgG) in paired samples from individuals who did not receive any COVID-19 vaccine following the pre-BA.1 sero-survey stratified by district, age-groups and vaccination status at pre-BA.1 sero-survey; Table S8: Seroconversion, sero-response and overall serological evidence of SARS-CoV-2 infection between the pre-BA.1 and post-BA.1 sero-surveys in individuals who did not receive any COVID-19 vaccine following the pre-BA.1 sero-survey and stratified by vaccination status at pre-Omicron sero-survey; Table S9: Seroconversion, sero-response and overall serological evidence of SARS-CoV-2 infection between the pre-BA.1 and post-BA.1 serosurveys in individuals who received any COVID-19 vaccine following the pre-BA.1 sero-survey and stratified by vaccination status at pre-BA.1 sero-survey; Table S10: Cumulative reported COVID-19 cases, hospitalizations, recorded deaths, and excess mortality in Gauteng Province by age-group and COVID-19 wave; Figure S1: Seroprevalence in pre-and post-Omicron serosurvey in unvaccinated individuals. Reference [10] is cited in Supplementary Materials. Institutional Review Board Statement: The Human Research Ethics Committee at the University of the Witwatersrand granted a waiver for ethics approval of the survey, which was conducted as part of public health surveillance by the Gauteng Department of Health.
Informed Consent Statement: All participants were, however, required to provide written informed consent, and individuals within a household were free to decline participation.

Data Availability Statement:
Requests for data sharing should be directed to Professor Shabir A. Madhi, email: shabir.madhi@wits.ac.za.